When you ask an AI chatbot a question, it builds an answer word by word. It picks whatever sounds right next, based on patterns it learned from reading a giant pile of text.
Most of the time, the answer is correct but sometimes it's not true. The chatbot wrote something that looks real, sounds real, and is wrong. People call this a hallucination or a fabrication.
The AI is not lying to you. It has no idea whether what it is writing is true or false. It is doing the same thing it always does: picking the next word that fits. Sometimes that lands on a fact. Sometimes it lands on a fiction. The voice does not change either way.
Here are three answers from the same AI chatbot, in the same conversation, in the same confident voice. For each one, decide: is the answer true or made up?
AI fabrications are not the result of one problem. You can think of the kinds of AI mistakes that cause hallucinations as four kinds of errors in a trenchcoat. Each lesson in this mini-module walks through one of these kinds of mistakes.
Long chats overflow. Facts you shared early on slide out of the AI's memory window, and it fills the gap with a guess.
The AI never learned about your topic. It cannot feel the gap, so it decorates it with pieces of other topics.
False stories get repeated online. The AI reads them all and retells them with the same confidence it uses for facts.
The AI was trained to agree with you. Wrong answers arrive friendly and sure, and it folds under pressure instead of defending the truth.
The first lesson shows what happens when a chat gets long enough to forget. Continue to Where did that message go?